std :: map clear（）在调试器中的性能？ [英] std::map clear() performance in debugger?

问题描述

附加的，微不足道的测试程序测试清空一个简单的std :: map的性能。使用MSVC 2008和2010，当从命令提示符执行时，调试构建将需要<30秒，但是当从调试器中执行时将近3分钟。调用clear（）是完全负责的区别。如果我进入调试器，callstack将总是指向HeapFree。

问题：为什么巨大的差异？我可以以某种方式更改调试堆设置，以便在调试器中执行时会很快。

The attached, trivial, test program tests the performance of emptying a simple std::map. Using MSVC 2008 and 2010, the debug build will take <30seconds when executed from a command prompt but almost 3 minutes when executed from within the debugger. The call to clear() is entirely responsible for the difference. If I break into the debugger the callstack will always point to HeapFree.
Question: Why the huge difference? Can I somehow change debug heap settings so it'll be fast when executed in the debugger?

#include <map>

int
main ( int, char )
{
    std::map< time_t, double > test;
    for ( int i = 0; i < 1000000; ++i )
    {
        test[i] = i / 3.14;
    }
    test.clear();
    return 0;
}

推荐答案

尝试设置环境变量_NO_DEBUG_HEAP = 1在程序的初始环境中。这会禁用Windows的内部调试堆，这可能会使调试内存损坏问题变得更加困难。

Trying setting the environment variable _NO_DEBUG_HEAP=1 in the program's initial environment. This disables Windows' internal debug heap, which might make it harder to debug memory corruption problems.

此 KB文章提到该标志，并且您可以推断，如果程序在没有该环境变量的调试器中运行，则禁用默认（低碎片堆）。另请参阅此博文，其中讨论了调试堆可以将他们的程序减慢3-5倍。

This KB article mentions the flag, and you can infer that the default (low-fragmentation heap) is disabled if the program is run in a debugger without that environment variable. See also this blog post, which discusses how the debug heap can slow down their program by 3-5 times.

这篇关于std :: map clear（）在调试器中的性能？的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！